GPU Hang and rcs0 reset on Cherrytrail platform
Hi,
We're seeing an issue on one of our Cherrytrail based devices since a long overdue upgrade that took us from a kernel based on 5.1 through to one based on 6.8 (both Fedora)
We get a GPU HANG in either our movie playback process or Xorg itself. We've managed to do a kernel bisect between 5.1 and 5.2 which shows the bug was introduced in bac24f59 - drm/i915/execlists: Enable coarse preemption boundaries for gen8
With this in mind, I managed to patch the latest kernel (6.10.0) with the aim of disabling pre-emption on chv - 0001-Disable-pre-emption-for-CHV.patch. It's quite probably not the way to go about this, I imagine either a quirk or a module parameter would be better here!
I'm currently soak testing the device with the patch so don't have a dmesg from when the error occured - but I can grab the i915 log messages from journalctl from a previous boot - i915-errors.log
Any thoughts on this? If this is a fix you'd like to include happy to work on any changes required.